Trying to understand the effects of varying inputs

[Laura-Alex]

Hello,

I am probably using megahit in a manner it wasn't designed for (on fasta sequences after previous sorting of metagenomic data by spacegraphcats. Also I'm using -r even though some of my reads are paired, because I can't yet figure it out). I'd like to understand it better. My issue is as follows. When my input is very small (say, 5 sequences out of which 4 cover my region of interest), I get the contig that I want (or at least I assume it's correct, because it's homologous to the other genes from my list). However, when I use it on a larger dataset which includes those same sequences, I no longer get the respective contig. It doesn't even seem to appear in the 'intermediate_contigs'. I suspect this might be because I am trying to recover uncommon paralogues of low-abundance organisms.

What is causing this, and is there a way to adjust the settings to do what I want? Help would be appreciated!